An RNA folds into its functional structure via a series of intermediate steps and, for each of them, chooses one out of many possible consecutive ones. Assigning a height, e.g. the free energy, to each point of the
interconnected state space a landscape unfolds in front of us, resembling plains, mountains, ridges, valleys and funnels. The latters can capture the RNA in a local minimum and thus stabilize a temporarily functional state for an extended period of time, e.g. allowing interaction with other regulatory RNAs or proteins. Prominent
examples can be found in the Hok-Sok system in Escherichia coli, the Tetrahymena group I intron and the FMN dependent riboswitch in Bacillus subtilis. To date, only a handful of computational methods exist to analyze the kinetics of RNA, and even fewer enable research on sufficiently large biologically relevant cases. We therefore propose to develop probabilistic algorithms and heuristics to tackle several key problems to understand and predict RNA folding kinetics.
On a mathematical level, the RNA folding process can be abstracted as a stochastic process whose states are alternative conformations for an RNA. The transitions correspond to local perturbations which transform a conformation into another. Thermodynamics studies assume convergence towards a Boltzmann equilibrium, in which various quantities can be efficiently computed. However, the time needed by an RNA to eventually reach its most stable state depends on its sequence, and may well exceed the lifetime of the molecule. Therefore, to get a more accurate view of RNA in a cellular context, one needs to study RNA kinetics, the evolution of the folding process along with time. Unfortunately, RNA kinetics studies turn out to be much more computationally
demanding than thermodynamics. Key problems, such as computing the energy required for reaching a state from another one are known to be computationally intractable, which explains the current absence of efficient and exact methods for predicting the kinetics of RNA, and motivates the design of efficient heuristics.
In this project, we will construct coarse-grain representative folding landscapes, consisting of macro- states via a novel structure sampling method, thereby lifting the length restrictions faced by exhaustive enumeration strategy.
Based on the simplified landscape, efficient methods to compute the energy barrier separating two neighboring states will be developed. From such representations, we will analyze the underlying folding kinetics as a Markov process. Such an analogy provides the evolution over time of the population densities for each macro-state. We will then adaptation these methods to changing energy landscapes, eventually allowing to study the folding kinetics of a growing RNA chain, i.e. co-transcriptional folding. The complete pipeline will be validated on a collection of occurrences of kinetic effects suspected on biological examples, including bacteria.
